EXECUTIVE PRIMER

Your Best Agent Is a Chatbot

Everything You Should Know About Chatbots
for Customer Engagement

1
TABLE OF CONTENTS
2 Introduction 16 Chapter 5
So You Want to Build You Own
3 Chapter 1
Chatbot 101 19 Chapter 6
Defining the Business Case
8 Chapter 2
Enterprise Chatbots in Action 21 Chapter 7
How Should Your Chatbot Look and Act?
10 Chapter 3
Not All Chatbots are Created Equal: The 24 Chapter 8
Intelligence Question Good Bots/Bad Bots, Pitfalls and Best
Practices
14 Chapter 4
Chatbots and Agents: Better Together 28 Next Steps

CHATBOT EXECUTIVE PRIMER 1

 I keep seeing you pop up
on web pages and on my
phone, what exactly are
you?
INTRODUCTION
Advances in artificial intelligence mean that chatbots
can automate more interactions than were previously
possible with older technology. Now enterprises can
use artificially intelligent chatbots as virtual agents or
assistants to replicate the effectiveness of their best
agents and reduce customer frustration and wait times.

I’m a chatbot, software that However, if you were to believe all the current hype
emulates humans in a chat about chatbots, you’d think that they’ll soon
conversation. Everyone’s
talking about us, even other • Be everywhere.
chatbots. Here’s why... • Take over every interaction.
• Learn on their own with no help from humans.
• Be so easy and fast to create that anyone can do it.

While some of this may be true for lightweight chatbots

that are built for fun in a developer’s spare time, it’s a
big stretch to apply this thinking to enterprise chatbots.

While technology has made major leaps in the decades

since chatbots were first developed, it’s not an easy,
do-it-yourself type of project for the enterprise. Read on
to get the scoop about what chatbots can do—and can’t
Chatbot Use On The Rise do—for the enterprise right now.
According to analyst firm Gartner, the usage of chatbots
(called virtual customer assistants by Gartner) will triple
through 2019 as enterprises seek to increase customer
satisfaction and reduce operating costs.

Source: “Seven Decision Points for Success with Virtual Customer

Assistants,” Gartner, July 26, 2016.

CHATBOT EXECUTIVE PRIMER 2

 1 WELCOME TO
CHATBOT 101

So what does a chatbot

do anyway?

I am an application that
emulates a human in a
conversation using a chat or
speech interface. If I were
emulating a human in a
contact center, you might call
me a virtual agent. Shall I tell
you more?

CHAPTER 1: CHATBOT 101 3

WHAT’S IN A NAME?
While chatbots have been around for a while, they are not all
the same. “Chatbot” refers to a broad range of technologies
that allow consumers to use a conversational interface to
accomplish tasks.
Lets begin by looking at a simple chatbot.
Simple Chatbot: Many popular chatbots are aimed at
casual use and are designed to deliver an entertaining
experience. If a simple chatbot doesn’t know the answer or
can’t understand the question, the stakes are relatively low
for the brand. For instance, we don’t really expect Apple’s
Siri or Amazon’s Alexa to know all the answers. These are
examples of “unbound” chatbots that have no specific
context to operate within. These are commonly known as
virtual assistants.
For businesses, chatbots need to be smarter to be effective.
“Intelligent chatbots” are able to integrate with enterprise
systems, leverage big data, and use artificial intelligence
to help customers resolve issues or conduct transactions.
These are also referred to as virtual agents or virtual
customer assistants.
These chatbots are being used by enterprises to achieve
incredible results.
CHAPTER 1: CHATBOT 101
Intelligent Chatbot: Some chatbots interact with
customers to resolve issues, conduct transactions, and
answer questions. The best enterprise chatbots mimic
the effectiveness of your top customer service agent,
salesperson, or support technician. The fact that these
chatbots are “bounded”, or operating within a certain context
(mortgages, utilities, wireless), ensures they can better
support the conversation. If these chatbots make a mistake,
the risk for the brand can be high. Consider what would
happen if a banking chatbot answered a policy question
inaccurately.
Where can you use chatbots?
You can deploy chatbots on nearly any interface (web,
mobile, social, messaging app, voice response, SMS).
How do they engage with customers?
Proactive: Intelligent chatbots can operate in real time and
predict customer intentions—offering specific help when they
detect that a customer may need assistance. For example,
a client has visited several mortgage pages and pauses on
a specific page whereby the chatbot can proactively engage
the client.
Reactive: Customers can select chatbots from a menu or a
button on a web page or in a mobile app. Other approaches
include chatbots as “listeners” (for example, twitter,
Facebook, or SMS) that react to inquiries as customers enter
these channels.
4
WHAT ARE THE STAGES OF CHATBOT MATURITY?

Informational Personalized Transactional

Understands natural language to provide Presents user-specific responses by Guides the user through a series of
answers to questions posed connecting to enterprise systems and steps to complete a task (can also
clarifies user intent using menus or be conversational) and can integrate
simple questions. customer data

Fig. 1 Chatbot Maturity Framework

Transactional
• Emulates human
conversation and
understands context to
Personalized
Contextual

• Presents user-specific
responses by connecting
UNDERSTANDING
complete transactions
• Executes transactions on
behalf of user

to enterprise systems

• Personalized responses • Handles complex intents

Informational
• Natural Language maps
user request to simple
Natural Language
include results from back-
end systems (e.g., CRM) • Expedites resolution
through proactive service
• Clarifies user intent using

intent menus or simple questions

• Intent maps to the best

response in the content
data base

Information Provision Transactional

ENGAGEMENT
CHAPTER 1: CHATBOT 101 5
WHAT ARE THE DRIVERS FOR DEPLOYING ENTERPRISE
CHATBOTS?
1 Changing customer expectations
Customers prefer the convenience and speed of on-
demand, accurate answers.
2 Decreasing customer satisfaction
Customers are dissatisfied with:
• Not finding the answers to their questions.
• Apps that take too much effort to download and use.
• Long wait times for agents.
3 Losing sales
Frustrated customers leave their shopping carts and go to
your competition when they don’t get the assistance they
need, when they need it.
CHAPTER 1: CHATBOT 101
4 Increasing volume
Chatbots are scalable, supporting hundreds of thousands
of conversations in parallel.
5 Reducing operating costs
Chatbots can divert customer interactions away from live
agents to reduce the cost of delivering services, allowing
live agents to focus on higher-value transactions.
6 Increasing visibility
Enterprises need insight into what customers are looking
for but not finding on their websites. They also need to
understand the changing trends based on customer
interactions. Enhanced analytics turns data from chatbot
conversations into actionable intelligence.
6
WHICH PARTS OF THE BUSINESS
CAN BENEFIT FROM CHATBOTS?

Sales and marketing

Chatbots can generate more leads, attract more customers, and increase
revenue from existing customers.
Chatbots are evolving
in ways that have a
Customer service and support
Chatbots can handle routine questions and tasks, freeing up more time for
profound effect on
staff to deal with more complex, value-added services. digital commerce and
the future of customer
Partner programs care and marketing.”
Chatbots can help partners get the information they need to support their
customers quickly and effectively.
– Dan Miller, Opus Research,
Enterprise Chatbots, Winning
Employee assistance Hearts, Minds, and Loyalty in the
Chatbots can: Digital Age

• Help train new employees.

• Provide human resources assistance.
• Assist contact center agents with questions.

Customer Insight
Chatbot conversations, often representing millions of interactions, create
the opportunity for cross-organizational analytics and insights. Marketing,
product, sales, and customer service benefit from a shared view of customer
trends as well as true voice of the customer insights.

CHAPTER 1: CHATBOT 101 7

 ENTERPRISE
2 CHATBOTS IN
ACTION

How are companies like

mine using chatbots
today?

Enterprise chatbots can

be extremely valuable
for companies in many
different industries. Here
are a few examples…

CHAPTER 2: ENTERPRISE CHATBOTS IN ACTION 8

CHATBOTS ACROSS INDUSTRIES
Communications
Telecommunications
company Vodafone’s award-
winning virtual agent “Hani”
is an intelligent chatbot
that answers 80,000
questions per month and
deflects calls away from
the contact center for 75
percent of the customers
it chats with. Vodafone
contact center staff also
use the same technology to
access accurate, up-to-date
information on Vodafone
products and services.
Healthcare
A major health insurance
provider improved the
experience for its 4 million
members with an intelligent
chatbot deployed as a
virtual agent. With the
chatbot answering 150,000
questions per month,
the company is saving
thousands of dollars in
contact center costs by
reducing calls to its staff.
CHAPTER 2: ENTERPRISE CHATBOTS IN ACTION
Travel
A leading global airline
created an avatar to
personify their chatbot.
The chatbot serves as
an automated concierge,
providing customers with
instant, accurate answers to
their questions about flight
status and baggage rules.
The chatbot has helped the
airline reduce call and chat
volume by 40 percent.
Utilities
Canadian utility BC Hydro
wanted to improve customer
service and satisfaction
for its 4 million customers
and improve operational
efficiency by deploying a
chatbot on its website. In
the first 11 months, the
chatbot answered more than
720,000 questions with an
accuracy rate of 94 percent.
Financial Services
Canadian Imperial Bank of
Commerce, one of Canada’s
largest chartered banks,
introduced an intelligent
chatbot as a virtual agent
and saw email volume
decrease by 50 percent
immediately at launch, and
then experienced another
23 percent drop throughout
the first year. At the same
time, it reduced phone calls
by 25 percent.
Retail
A major retailer
implemented an intelligent
chatbot to deliver a
phenomenal guest
experience, answering
45,000 questions a
month about order status,
shipping, returns, and other
common areas of interest.
The chatbot deflects
informational calls and
email away from staff by
answering 97 percent of the
questions asked, with 96
percent accuracy.
CHATBOTS
DELIVER
RESULTS
Deploying an enterprise
chatbot can help your
company:
• Increase customer self-
service engagement.
• Improve customer
satisfaction ratings, lower
customer effort scores,
and increase your Net
Promoter Score.
• Automate routine
customer questions to
allow human agents to
focus on higher-value
interactions.
• Deflect calls, email, and
chats to reduce costs.
• Create a seamless hand-
off from self-service virtual
assistance to a live agent.
• Generate true “voice
of the customer” data
through the conversations.
9
 NOT ALL CHATBOTS
ARE CREATED EQUAL:
3
THE INTELLIGENCE
QUESTION

What makes you so

smart, chatbot?

Thank you very much.

Unlike others, I’m an
intelligent chatbot. Here’s
what that means…

CHAPTER 3: NOT ALL CHATBOTS ARE CREATED EQUAL 10

 You won’t get significant operational
savings or big leaps in customer
satisfaction with a lightweight,
simplistic chatbot. For these kinds
of results, you need an intelligent,
enterprise-ready chatbot, one that is
built to take advantage of:
1. All your data
Data from across your enterprise helps predict customer
needs and deepen customer understanding. You should also
use it to unify the customer journey by adding context across
interactions, even as customers cross channels or move
from chatbot to a live agent.
2. Sophisticated natural language
processing (NLP)
Artificial intelligence (AI) is a collection of abilities that
machines exhibit that we normally associate with human
intelligence such as NLP. This part of AI focuses on
categorizing and understanding the language between
human and machine. NLP is the ability to process free-form
text entered by customers to determine expressed intent
(i.e., intent that is conveyed by language). Machine Learning
models are trained on tens of thousands of customer
messages to classify requests and extract query parameters.
CHAPTER 3: NOT ALL CHATBOTS ARE CREATED EQUAL
3. Advanced AI and machine learning
Intelligent chatbots use AI to make sense of data and context
in real time to quickly learn and become more effective, this
is known as machine learning. Machine learning enables
AI by processing and analyzing large amounts of data.
The more data used for machine learning, the better. An
intelligent chatbot uses machine learning to continuously
become smarter by learning from conversations it has with
people.
4. Advanced analytics and insights
Intelligent chatbots derive considerable hard data around
intent. The ability, for example, to analyze what the chatbot
didn’t understand can lead to valuable new product insights.
Analyzing the conversations happening across chatbots
(thousands or even millions of conversations) allows
organizations to understand how the community asks
questions and works to resolve issues. This insight drives
next issue avoidance (where the chatbot knows what the
customer likely needs to know next, based on the current
point of the conversation). Appearing and disappearing
trends analyzed day-to-day, monthly, or seasonally provide
insights into shifting needs and sentiments.
11

HERE’S HOW AN
INTELLIGENT CHATBOT
WORKS...
1. Captures data in real time
The intelligent chatbot captures the customer’s identity, attributes,
and engagement data, and any feedback the customer provides—
all in real time.
For example, the chatbot determines:
• Date, time, physical location, and device information
• Whether the customer is on the web or a mobile app
• Whether the customer requested to engage with a chatbot or received
a proactive invitation
• Where the customer was on the website or mobile app when he or
she began the interaction with the chatbot
2. Uses internal data
Using data such as customer profile and preferences, value to the
company, location, industry, and amount of money spent in the
past year gives the chatbot more insights about the customer. This
data is gathered from various sources and is typically available in
customer relationship management (CRM) systems.
3. Combines data to predict customer
intentions
The chatbot develops an understanding of what the customer
wants/needs by combining all the data signals. This helps make
the conversation contextual and more natural when the customer
engages the chatbot.
CHAPTER 3: NOT ALL CHATBOTS ARE CREATED EQUAL
4. Engages customers
Customers can invoke chatbots themselves when they need
assistance, or chatbots can proactively engage customers.
5. Understands what is said
The chatbot takes each message written or each utterance
spoken and runs it through natural language models to
understand what the customer said. This interaction is contextual
and personalized to the customer. The chatbot achieves this by
leveraging information such as the web page the customer was
on when they engaged with the chatbot and their customer profile.
For example, if a customer is on a bank’s website looking at a
page on mortgages and asks the chatbot what the interest rate is,
the chatbot will know the customer is asking about the interest rate
for mortgages.
6. Formulates a response
Once the chatbot understands the customer’s intent, the response-
matching algorithm determines the correct response and
assembles it from knowledge bases and CRM systems.
7. Determines follow-up actions
If the customer provides feedback that he or she is satisfied with
the chatbot response, the chatbot closes that intent and waits for
a new intent. If the customer requests the chatbot to help “pay
my credit card bill,” for example, the chatbot will determine the
appropriate follow-up actions such as asking the customer for a
password and then completing the transaction.
WHAT MAKES A CHATBOT ENTERPRISE-READY?
1 Security and Authentication
Security and authentication are top priorities in heavily
regulated industries such as financial services and
healthcare.
4 Scalability
Scalability lets you handle exponentially and simultaneously
more interactions as your chatbot becomes more popular
with customers.

2 Reliability 5 Compliance
Reliability is critical so customers aren’t disappointed by a Compliance with industry standards such as HIPAA, PCI,
chatbot that isn’t available or working correctly. and SOX further protects the privacy and security of your
customer data.

3 Prebuilt Models 6 Guided Learning

Prebuilt Models in your industry accelerates your chatbot’s Guided Learning ensures that the chatbot can be trained
knowledge and lets you achieve your goals more quickly. to understand new concepts or have learning reinforced
through community feedback (perhaps your internal agents
or customer feedback).

Virtual agents have evolved significantly over the last few years. They are now
powerful, big-data-driven solutions that can connect to back-end architecture
and deliver results from nearly any other system in the enterprise.”
– Brendan Witcher, Forrester, Get Better Results - Virtual Agents Are Key To Improving Site Search

CHAPTER 3: NOT ALL CHATBOTS ARE CREATED EQUAL 13

 CHATBOTS AND
4 AGENTS: BETTER
TOGETHER

Will chatbots replace

my agents?

No, at least not in the

near future. But we do
complement each other.

CHAPTER 4: CHATBOTS AND AGENTS 14

 Intelligent chatbots help customer-care organizations work more efficiently by:

Minimizing menial or repetitive Mining agent interactions to learn Reducing average handling time Maintaining the context of
work for agents by deflecting new customer intents and by suggesting responses while the previous interactions, thus
interactions. agent solutions. agent is chatting with the customer. avoiding “starting over”.

The first rule of any technology used in a business is that automation applied to
an efficient operation will magnify the efficiency. The second is that automation
applied to an inefficient operation will magnify the inefficiency.”
- Bill Gates

When is a human 1
1. The customer’s request is not 4 The customer’s request is better served
4.
required? understandable. by an agent (e.g., conversion or attrition).

2 5 It is a high-value transaction and your

A chatbot should
2. The customer appears to be annoyed or 5.
frustrated. company wants a live agent to close the
escalate to a live sales opportunity.
agent when: 3
3. The customer’s request cannot be
handled in self-service (due to rules or 6 The customer explicitly requests a
6.
policies). human agent.

4: CHATBOTS AND AGENTS 15

 I’m hearing it’s easy to
build your own chatbot.
Is that true?

Yes and no. It’s fairly

straightforward to build a
simple chatbot, but that may
not be what you need.
Here’s why…

SO YOU WANT TO
5
BUILD YOUR OWN
CHAPTER 5: SO YOU WANT TO BUILD YOUR OWN 16

WHAT ARE THE
DISADVANTAGES OF
TRYING TO BUILD YOUR
OWN CHATBOT?
1. Not as intelligent
Do-it-yourself (DIY) chatbots typically don’t employ machine
and self-learning, and many don’t understand true intent. For
instance, a simple DIY chatbot might not know that canceling a
check is not a bad thing, but canceling a service can be.
2. Limited scope
An intelligent chatbot from an experienced chatbot provider
can handle hundreds of different requests out of the box. It’s
been industry trained across millions of interactions, while DIY
chatbots require the user to typically build the understanding
and intent from scratch.
3. Higher risk of failure
If the chatbot doesn’t help customers, they aren’t likely to use it
in the future. Worse, what if your chatbot “misbehaves,” such as
providing incorrect information? For a consumer experience that
is meant to be entertaining, an incorrect answer doesn’t carry
nearly as much cost and risk as it does in the enterprise.
CHAPTER 5: SO YOU WANT TO BUILD YOUR OWN
4. Expensive custom development
While there are chatbot tools and smart-machine platforms (e.g.,
IBM Watson, Microsoft Bot Framework, and Facebook
Wit.ai), these platforms are primarily used for general purposes.
Deploying them in the enterprise will require extensive
development efforts and big budgets.
5. Integration burden
Enterprise chatbots from a proven provider often deliver ready-
made integration with popular enterprise information systems
as well as standardized interfaces for integration with custom
applications. But you must integrate DIY chatbots from scratch.
6. Longer time to deploy
It will take your team much longer to build a natural-language
model and a knowledge/context base when you are starting
from scratch.
7. Higher costs
Custom development and longer development cycles can
negate perceived cost benefits of DIY chatbots versus proven
chatbot solutions.
8. Lack of analytics
Typically, DIY platforms focus only on the “optics”. One of
the true benefits extracted from chatbots is the ability to
meaningfully analyze the conversations to isolate key trends and
understandings.
17
Comparing Simple vs. Intelligent Chatbots
Simplistic chatbots use logic that is based on keywords or sentences and perform simple pattern matching. These bots can only respond to
very specific commands. If users say the wrong thing, the basic chatbot won’t know what they mean.

It’s far more difficult to build an intelligent enterprise chatbot that understands and responds intelligently to the true meaning of the words
and phrases that users enter.

FEATURE/CAPABILITY SIMPLE/DIY CHATBOT INTELLIGENT CHATBOT

Reads natural language Predefined NL understanding and

grammar processing

Applies logic

Learns through experience

Manual efforts Machine learning

Understands intent Explicit keyword Big data-driven

spotting and scoring understanding
Escalates to a human agent when it can’t answer a question, address an issue, or
complete a task
Understands context

Applies context to distinguish between different uses of the same word to understand the
implied meaning
Uses follow-up questions to clarify intent

Understands, learns, and interacts as a human would to solve problems

Manages tasks and completes transaction

Accesses enterprise information to personalize the experience or perform transactions

Independently assesses what actions it needs to execute at each step of an overall

process to resolve the problem as swiftly as possible

CHAPTER 5: SO YOU WANT TO BUILD YOUR OWN 18

 How can a chatbot help
my business? Will it save
us money?

Yes, it definitely can. In

fact, intelligent chatbots
deliver tangible business
value in a number of
ways. But before we get
into that, it’s important to
decide what metrics you
are trying to improve.
Here are some ideas…

BUSINESS CASE
6
FOR CHATBOTS

CHAPTER 6: BUSINESS CASE FOR CHATBOTS 19

DEFINING YOUR BUSINESS CASE
Defining your business case starts with understanding your desired outcomes for NOW THAT’S A
deploying an enterprise chatbot. For instance, you might want to reduce call volumes BUSINESS CASE
in your contact center by as much as 50 percent. Maybe you also want to reduce
shopping cart abandonment by 10 percent. For enterprises, these outcomes could A North American retailer wanted to
mean tens of millions of dollars in cost reduction and millions in increased revenue. increase incremental revenue for every
online transaction, improve efficiency,
Here are some ways to predict and measure a chatbot’s return on investment as part drive higher self-service rates, and
of your business case: provide consistently delightful and
engaging experiences across all
channels.
1 Reducing costs
Using an intelligent chatbot over a six-
Deflecting live chats, email, and phone support through self-service month period, the retailer:
automation reduces operational costs for your contact center.
• Kept 55 percent of contacts within
2 Supporting growth without increasing costs the web self-service channel.
Customers could solve their
By handling more customer interactions with a chatbot and deflecting the problems without speaking with a
number of interactions that require a live agent, you can grow your customer human agent or making a phone
base, expand your geographic footprint, and introduce new products and call.
services without increasing your human capital costs.
• Saved $1.6 million in contact

Monetizing self-service
center costs.
3
A chatbot gives you the opportunity to build and grow new revenue • Received $27.8 million in chat-
streams through proactive offer management and reduced shopping cart contributed revenue contacts within
abandonment. the web self-service channel,
meaning that customers could
solve their problems without
4 Improving customer loyalty speaking with a human agent or
Reducing customer effort keeps customers coming back, which in turn, making a phone call.
keeps your profitability higher as your customer lifetime value increases.
Happy customers also refer their friends to your brand.

CHAPTER 6: BUSINESS CASE FOR CHATBOTS 20

 What characteristics
should my enterprise
chatbot have?

That will depend, in large

part, on your company’s
brand and the industry
you compete in, but here
are a few things to keep
in mind…

HOW SHOULD
7 YOUR CHATBOT
LOOK AND ACT?
CHAPTER 7: HOW SHOULD YOUR CHATBOT LOOK AND ACT? 21
 AVATARS
Giving your chatbot a name like “Pam”, along with a visual
persona or avatar, can increase your chatbot’s activation
and engagement rates as well as enhance your brand image
(much like Siri does for Apple). However, if you decide to
use an avatar for your chatbot, be wary of potential pitfalls.
When anthropomorphizing your chatbot, make sure you aren’t
unintentionally offending any customers, particularly if your
company is global. Additionally, avatars don’t display well on
mobile devices where screen real estate is at a premium.

Fig. 3 Chatbot avatar

ON MOBILE
You need to be where your customers
are, and increasingly they are using
mobile devices to interact with your
enterprise. Chatbots need to be mobile-
ready, that is, designed for optimal use
on a small screen.

Fig. 4 Mobile chatbot

CHAPTER 7: HOW SHOULD YOUR CHATBOT LOOK AND ACT? 22

 BRAND
Your chatbot should represent your
corporate or brand style; otherwise,
you risk confusing your customers
on what they can expect from your
company.

Fig. 5 Chatbot representing brand style

CONVERSATION STYLE
Match the style of your chatbot’s conversation
to your business. For example, a brokerage or
bank would have a more formal conversational
style than a brand aimed at youthful customers.

Fig. 6 Conversation style aligning with company

CHAPTER 7: HOW SHOULD YOUR CHATBOT LOOK AND ACT? 23

 I know all chatbots aren’t
the same. What makes a
chatbot good or bad?

Bad bots hurt your brand because they

don’t offer much help to customers.
Good bots converse with your
customers to help them get what they
want quickly. They have the intelligence
and integration to solve problems while
also being able to escalate to a live chat
agent when necessary.

GOOD BOT/BAD BOT:

8 PITFALLS AND BEST
PRACTICES
CHAPTER 8: GOOD BOT/BAD BOT 24
HOW DO YOU AVOID
DEPLOYING A BAD BOT?
Don’t deploy a bot on its own
Chatbots need to be part of the entire customer-
engagement strategy; otherwise, it can create a
disjointed, siloed, or dead-end experience for the
customer.
Don’t pretend it’s a human
Make sure customers don’t think they’re chatting with a
real person only to find out they’re really interacting with
a chatbot. This will create misaligned expectations.
Always escalate at the right time
If customers aren’t getting the answers they need, you
risk frustrating and alienating them if they can’t quickly
speak to a person.
CHAPTER 8: GOOD BOT/BAD BOT
Keep it relevant
Chatbots aren’t set-and-forget solutions. You need
to plan for the right amount of resources for content
maintenance and regular chatbot learning.
Don’t hide the chatbot
Design, placement, and branding of the chatbot are
paramount so customers can find it and engage with it
easily. The chatbot must be part of the entire website or
mobile experience. True benefits arise when the chatbot
is proactive in identifying opportunities to engage.
Collaborate across departments
The marketing, digital experience, sales, and customer
care departments need to collaborate so customers have
a seamless experience.
25

WHAT ARE THE BEST
PRACTICES FOR
DEPLOYING GOOD
CHATBOTS?
1. Respect form factors: Especially with mobile technology,
there is limited “real estate” to create your chatbot experience.
Insist on a clean design that honors the customer’s expectations
about how and where the conversation will unfold.
2. Integrate with enterprise information systems: Effective
agents (both live and virtual) need access to enterprise systems
and data to resolve customer questions and issues.
3. Stick to chat-like interfaces: Customers know how chat
works; so follow general chat best practices to make the
experience intuitive and easy for customers to use.
4. Make it natural: Focus on giving your chatbot a
conversational tone, one that is unassuming but with some
emotive aspects to frame the conversation.
5. Be contextual and to the point: Customers need help and
may be impatient, confused, or overwhelmed at the moment of
interaction with your chatbot. Your chatbot should tell them what
it knows about them and draw from available data points to
make the experience as efficient as possible.
CHAPTER 8: GOOD BOT/BAD BOT
6. Align with your brand: When designing your chatbot, make
sure that it looks like a natural part of your brand’s experience.
Follow your company’s standardized color palette, fonts,
imagery, and other aspects of your brand identity.
7. Use first person: It’s a one-to-one conversation, so your
chatbot should speak in singular, first person “I.” When referring
to the brand or team, it’s OK to use “we.”
8. Make it succinct: When it comes to information or
conversation, customers don’t want pages of monologue. Send
several messages, not one big paragraph. Make it easy to scan
and consume.
9. Don’t ignore the humans: While AI technology has
advanced, we’re still at the beginning of its evolution. Even
intelligent chatbots aren’t totally self-learning. You still need
humans to supervise their learning, analyze and prepare training
data, and integrate chatbots with enterprise information systems.
10. Insist on out-of-the-box industry knowledge: Look for
a chatbot solution proven for your industry, with out-of-the-box
intelligence about industry practices, terminology, and customer
journeys. This will help you automate 20 percent of customer
journeys right away; then apply AI and machine learning to
achieve a further 50 percent or more automation.
11. Make maximum use of chatbot conversations: Mine
the hundreds of thousands of conversations your chatbot is
having in real time to judge customer sentiment, drive increased
customer understanding, create new products, and adapt to
changing customer needs. A good enterprise chatbot does
more than converse; it unlocks the intent and understanding that
comes from analyzing conversations.
26
 Conversation
CHECKLIST FOR WHAT
Your chatbot should enable you to personalize content and
TO LOOK FOR IN A orchestrate experiences that continue the conversation
across channels and devices. You need the ability to view the
CHATBOT SOLUTION conversation as part of a journey, using contextual information
from what happened before to improve the current interaction.

Purpose-built
Choose a chatbot designed for solving your customer service Learning
issues, not a general solution.
Combine agent chat and other interactions with chatbot
interactions to improve learning for your chatbot. Tag and label
conversations and use them to train chatbots.

Extensive industry experience

A chatbot that already has a deep knowledge of your industry
saves you time and money because it already has domain
understanding. It only needs to augment this understanding with
your company’s specific dialog.
Data control
Escalation
You need the ability to escalate or redirect to an agent. The
transition must be seamless. Agents need to see the customer’s
interaction with the chatbot to speed up the resolution and
reduce customer frustration. Having context creates a seamless
customer experience.

You need the ability to combine data from the entire customer
life cycle—before acquisition through customer engagement.
Data analysis
You should be able to mine customer interactions for assisted
Intent prediction or guided learning for building chatbot intelligence and to gain
further customer insights.
Your chatbot needs to predict customer intent based on
observing customer behaviors over millions of interactions.

CHAPTER 8: GOOD BOT/BAD BOT 27

 NEXT STEPS
A pioneer in AI for customer engagement, [24]7 has deployed more enterprise chatbots than any other vendor on the market
today. Intelligent chatbots from [24]7 provide real business results and proven return on investment.

Offering the technology, people, and processes to deploy chatbots quickly and effectively across most industries, [24]7 is
redefining the way that companies interact with consumers to improve digital adoption, customer satisfaction, and revenue
growth.

Experience
[24]7 has been providing chatbots for more than 15 years and we have more than 500 deployments under our belt. The
company’s solutions assist several hundred million visitors across all channels, and engages in more than 1.5 billion
conversations annually, most of which are automated.

To see [24]7 Chatbots in action, request a demo today.

Let [24]7 help your organization achieve About [24]7

extraordinary results. [24]7 is redefining customer acquisition and engagement by
Contact us today. making consumer intent the cornerstone of digital transformation.
With intent-driven engagement, companies anticipate and act on
www.247-inc.com consumer intent across any channel, collapsing the time to deliver
successful outcomes in the moments that matter most.
queries@247-inc.com
USA +1 855.692.9247 For more information visit: www.247-inc.com
CA +1 866.454.0084
UK +44 0.207.836.9203
AUS +61 2.90025780 1702EBCHATBOT
©2017 24/7 Customer, Inc. All Rights Reserved.28